README.md

---
license: apache-2.0
language:
- en
- zh
base_model:
- deepseek-ai/DeepSeek-R1-Distill-Qwen-7B
pipeline_tag: text2text-generation
tags:
- Reward_Model
- Reasoning_Model
---
# Model Card for Model ID

## Model Details

### Model Description

<!-- Provide a longer summary of what this model is. -->

- **Developed by:** Hao Peng@THUKEG
- **Model type:** Generative reward model
- **Language(s) (NLP):** English, CHinese
- **License:** apache-2.0
- **Finetuned from model [optional]:** deepseek-ai/DeepSeek-R1-Distill-Qwen-7B


### Model Sources [optional]

<!-- Provide the basic links for the model. -->

- **Repository:** https://github.com/THU-KEG/VerIF
- **Paper:** https://arxiv.org/abs/2506.09942


## Training Details

### Training Data

<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

This model is trained from DeepSeek-R1-Distill-Qwen-7B using 131k critic data [IF-Verifier-Data](https://huggingface.co/datasets/THU-KEG/IF-Verifier-Data). 
This model is used for verifying soft constraints of instruction following.
Deploying IF-Verifier-7B requires only one single H800 GPU, with an average reward computation time of **120** seconds per batch, which can be further reduced with multi-GPUs.

### Results
The model trained using this model is comparable with that of QwQ 32B.

![Result fig](result.png)


#### Summary
Please refer to our paper and our GitHub repo (https://github.com/THU-KEG/VerIF) for more details.

## Citation
If this model helps, please kindly cite us:
```
@misc{peng2025verif,
      title={VerIF: Verification Engineering for Reinforcement Learning in Instruction Following}, 
      author={Hao Peng and Yunjia Qi and Xiaozhi Wang and Bin Xu and Lei Hou and Juanzi Li},
      year={2025},
      eprint={2506.09942},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2506.09942}, 
}
```